Low-frequency Validation of Hybrid Earth Models in the Salton Trough, Southern California and Significance for Urban Seismology

Abstract:

A potential seismic hazard facing Southern California is a large-magnitude earthquake from a San Andreas fault rupture with energy propagating into the densely populated greater Los Angeles area. To predict the expected ground-shaking from such an event accurately, researchers have put significant effort into developing seismic velocity models for the area. The goal is to use the ground motion estimates and structural engineering information to help better prepare Southern Californians. In this research, we present our evaluation of Earth models in the vicinity of the southern San Andreas fault in the Salton Trough. We use ground motion from earthquakes recorded along the Salton Trough axis with magnitudes between 3.5 and 5.5 that have simple source descriptions to evaluate the models to a maximum frequency of 0.5 Hz. The suite of Earth models we validate includes the popular regional Southern California Earthquake Center (SCEC) CVM-H 15.1 and CVM-S4.26 models and hybrid models constructed by embedding hi-res basin models developed using active-source data into the regional SCEC models. By generalizing the definition of typical window functions such as Cosine and Ormsby filters to arbitrarily shaped domains, we can smoothly embed hi-res basin models in non-rectangular support volumes while using a regular grid data structure. We implemented our algorithm into the SCEC Unified Community Velocity Model (UCVM) software, thereby extending the program's functionality to construct complex hybrid earth models rapidly. Before embedding the basin models, we benchmark the community models by validating several representations that include topography, attenuating rheology, and a geotechnical layer or the lack thereof. For the CVM-S4.26 model developed without topography, we experiment with 1-D and linearly interpolated model extensions from sea level and a pull-up parameterization that maps the sea level model boundary upward to represent the real topography. The final model assessment is done by computing the waveform difference in the observed and predicted data using frequency-dependent squared error normalized by the geometric mean of the energy in the waveforms. Our findings indicate that additional effort is needed to update the regional models to higher frequencies of interest in structural engineering.